Closed farley13 closed 7 years ago
One comment here - if possible we should run all of these on 1 VM with the load balancer pointing to it - this hopefully is the most cost effective way of dealing with this. We can deal with some down time when we release until we have more traffic.
If I've got it correctly. There is no point to use auto-scale group for Redis and Solr.
In fact auto-scale group is clusterization technology. However, Redis use their own cluster (see redis-trib.rb
). If we will want to scale Solr horizontally, we should use its own sharding (e.g. ZooKeeper)
"...if possible we should run all of these on 1 VM..." Yes, it is possible to run all of the our containers: PostrgreSQL, Solr and all Redis nodes on 1 VM. To be more precise 1 VM for TEST and 1 VM for PROD.
I've never tried t2.nano. Should I try it or t2.micro is fine?
Let's try nano and if it's totally broken move up to micro?
Creating a new issue for the auto-redeploy of updated configuration/auto-recreation if the current VM goes down (my understanding would be that we would lose the DB changes and we'd need to manually spin a new VM up).
Currently Jenkins is only deploying the web app to an auto-scale group - we should look to do the same for redis, solr. We likely should start with a single instance deploy for both for now (autoscale -> just 1 instance) - similar to the dev setup. The only critical thing here is ensuring we have actual uptime for both when AWS EAST inevitably goes down. It should spin up a new instance of either automatically if the old one is killed.